blob: 095bc453ff4c29e0f8ae7b19a2017b8d4fa0c731 [file] [log] [blame]
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001/*
2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
4 *
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
8 * 6WIND, Paris, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 */
18
19#include <asm/system.h>
20#include <asm/uaccess.h>
21#include <linux/types.h>
22#include <linux/sched.h>
23#include <linux/errno.h>
24#include <linux/timer.h>
25#include <linux/mm.h>
26#include <linux/kernel.h>
27#include <linux/fcntl.h>
28#include <linux/stat.h>
29#include <linux/socket.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090030#include <linux/inet.h>
31#include <linux/netdevice.h>
32#include <linux/inetdevice.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090033#include <linux/proc_fs.h>
34#include <linux/seq_file.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090035#include <linux/init.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090036#include <net/protocol.h>
37#include <linux/skbuff.h>
38#include <net/sock.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090039#include <net/raw.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090040#include <linux/notifier.h>
41#include <linux/if_arp.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090042#include <net/checksum.h>
43#include <net/netlink.h>
44
45#include <net/ipv6.h>
46#include <net/ip6_route.h>
47#include <linux/mroute6.h>
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090048#include <linux/pim.h>
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090049#include <net/addrconf.h>
50#include <linux/netfilter_ipv6.h>
51
52struct sock *mroute6_socket;
53
54
55/* Big lock, protecting vif table, mrt cache and mroute socket state.
56 Note that the changes are semaphored via rtnl_lock.
57 */
58
59static DEFINE_RWLOCK(mrt_lock);
60
61/*
62 * Multicast router control variables
63 */
64
65static struct mif_device vif6_table[MAXMIFS]; /* Devices */
66static int maxvif;
67
68#define MIF_EXISTS(idx) (vif6_table[idx].dev != NULL)
69
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090070static int mroute_do_assert; /* Set in PIM assert */
71#ifdef CONFIG_IPV6_PIMSM_V2
72static int mroute_do_pim;
73#else
74#define mroute_do_pim 0
75#endif
76
Rami Rosen6ac7eb02008-04-10 12:40:10 +030077static struct mfc6_cache *mfc6_cache_array[MFC6_LINES]; /* Forwarding cache */
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090078
79static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */
80static atomic_t cache_resolve_queue_len; /* Size of unresolved */
81
82/* Special spinlock for queue of unresolved entries */
83static DEFINE_SPINLOCK(mfc_unres_lock);
84
85/* We return to original Alan's scheme. Hash table of resolved
86 entries is changed only in process context and protected
87 with weak lock mrt_lock. Queue of unresolved entries is protected
88 with strong spinlock mfc_unres_lock.
89
90 In this case data path is free of exclusive locks at all.
91 */
92
93static struct kmem_cache *mrt_cachep __read_mostly;
94
95static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache);
Rami Rosen6ac7eb02008-04-10 12:40:10 +030096static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +090097static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm);
98
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +090099#ifdef CONFIG_IPV6_PIMSM_V2
100static struct inet6_protocol pim6_protocol;
101#endif
102
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900103static struct timer_list ipmr_expire_timer;
104
105
106#ifdef CONFIG_PROC_FS
107
108struct ipmr_mfc_iter {
109 struct mfc6_cache **cache;
110 int ct;
111};
112
113
114static struct mfc6_cache *ipmr_mfc_seq_idx(struct ipmr_mfc_iter *it, loff_t pos)
115{
116 struct mfc6_cache *mfc;
117
118 it->cache = mfc6_cache_array;
119 read_lock(&mrt_lock);
120 for (it->ct = 0; it->ct < ARRAY_SIZE(mfc6_cache_array); it->ct++)
121 for (mfc = mfc6_cache_array[it->ct]; mfc; mfc = mfc->next)
122 if (pos-- == 0)
123 return mfc;
124 read_unlock(&mrt_lock);
125
126 it->cache = &mfc_unres_queue;
127 spin_lock_bh(&mfc_unres_lock);
128 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
129 if (pos-- == 0)
130 return mfc;
131 spin_unlock_bh(&mfc_unres_lock);
132
133 it->cache = NULL;
134 return NULL;
135}
136
137
138
139
140/*
141 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
142 */
143
144struct ipmr_vif_iter {
145 int ct;
146};
147
148static struct mif_device *ip6mr_vif_seq_idx(struct ipmr_vif_iter *iter,
149 loff_t pos)
150{
151 for (iter->ct = 0; iter->ct < maxvif; ++iter->ct) {
152 if (!MIF_EXISTS(iter->ct))
153 continue;
154 if (pos-- == 0)
155 return &vif6_table[iter->ct];
156 }
157 return NULL;
158}
159
160static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
161 __acquires(mrt_lock)
162{
163 read_lock(&mrt_lock);
164 return (*pos ? ip6mr_vif_seq_idx(seq->private, *pos - 1)
165 : SEQ_START_TOKEN);
166}
167
168static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
169{
170 struct ipmr_vif_iter *iter = seq->private;
171
172 ++*pos;
173 if (v == SEQ_START_TOKEN)
174 return ip6mr_vif_seq_idx(iter, 0);
175
176 while (++iter->ct < maxvif) {
177 if (!MIF_EXISTS(iter->ct))
178 continue;
179 return &vif6_table[iter->ct];
180 }
181 return NULL;
182}
183
184static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
185 __releases(mrt_lock)
186{
187 read_unlock(&mrt_lock);
188}
189
190static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
191{
192 if (v == SEQ_START_TOKEN) {
193 seq_puts(seq,
194 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
195 } else {
196 const struct mif_device *vif = v;
197 const char *name = vif->dev ? vif->dev->name : "none";
198
199 seq_printf(seq,
Al Virod430a222008-06-02 10:59:02 +0100200 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900201 vif - vif6_table,
202 name, vif->bytes_in, vif->pkt_in,
203 vif->bytes_out, vif->pkt_out,
204 vif->flags);
205 }
206 return 0;
207}
208
209static struct seq_operations ip6mr_vif_seq_ops = {
210 .start = ip6mr_vif_seq_start,
211 .next = ip6mr_vif_seq_next,
212 .stop = ip6mr_vif_seq_stop,
213 .show = ip6mr_vif_seq_show,
214};
215
216static int ip6mr_vif_open(struct inode *inode, struct file *file)
217{
218 return seq_open_private(file, &ip6mr_vif_seq_ops,
219 sizeof(struct ipmr_vif_iter));
220}
221
222static struct file_operations ip6mr_vif_fops = {
223 .owner = THIS_MODULE,
224 .open = ip6mr_vif_open,
225 .read = seq_read,
226 .llseek = seq_lseek,
227 .release = seq_release,
228};
229
230static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
231{
232 return (*pos ? ipmr_mfc_seq_idx(seq->private, *pos - 1)
233 : SEQ_START_TOKEN);
234}
235
236static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
237{
238 struct mfc6_cache *mfc = v;
239 struct ipmr_mfc_iter *it = seq->private;
240
241 ++*pos;
242
243 if (v == SEQ_START_TOKEN)
244 return ipmr_mfc_seq_idx(seq->private, 0);
245
246 if (mfc->next)
247 return mfc->next;
248
249 if (it->cache == &mfc_unres_queue)
250 goto end_of_list;
251
252 BUG_ON(it->cache != mfc6_cache_array);
253
254 while (++it->ct < ARRAY_SIZE(mfc6_cache_array)) {
255 mfc = mfc6_cache_array[it->ct];
256 if (mfc)
257 return mfc;
258 }
259
260 /* exhausted cache_array, show unresolved */
261 read_unlock(&mrt_lock);
262 it->cache = &mfc_unres_queue;
263 it->ct = 0;
264
265 spin_lock_bh(&mfc_unres_lock);
266 mfc = mfc_unres_queue;
267 if (mfc)
268 return mfc;
269
270 end_of_list:
271 spin_unlock_bh(&mfc_unres_lock);
272 it->cache = NULL;
273
274 return NULL;
275}
276
277static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
278{
279 struct ipmr_mfc_iter *it = seq->private;
280
281 if (it->cache == &mfc_unres_queue)
282 spin_unlock_bh(&mfc_unres_lock);
283 else if (it->cache == mfc6_cache_array)
284 read_unlock(&mrt_lock);
285}
286
287static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
288{
289 int n;
290
291 if (v == SEQ_START_TOKEN) {
292 seq_puts(seq,
293 "Group "
294 "Origin "
295 "Iif Pkts Bytes Wrong Oifs\n");
296 } else {
297 const struct mfc6_cache *mfc = v;
298 const struct ipmr_mfc_iter *it = seq->private;
299
300 seq_printf(seq,
301 NIP6_FMT " " NIP6_FMT " %-3d %8ld %8ld %8ld",
302 NIP6(mfc->mf6c_mcastgrp), NIP6(mfc->mf6c_origin),
303 mfc->mf6c_parent,
304 mfc->mfc_un.res.pkt,
305 mfc->mfc_un.res.bytes,
306 mfc->mfc_un.res.wrong_if);
307
308 if (it->cache != &mfc_unres_queue) {
309 for (n = mfc->mfc_un.res.minvif;
310 n < mfc->mfc_un.res.maxvif; n++) {
311 if (MIF_EXISTS(n) &&
312 mfc->mfc_un.res.ttls[n] < 255)
313 seq_printf(seq,
314 " %2d:%-3d",
315 n, mfc->mfc_un.res.ttls[n]);
316 }
317 }
318 seq_putc(seq, '\n');
319 }
320 return 0;
321}
322
323static struct seq_operations ipmr_mfc_seq_ops = {
324 .start = ipmr_mfc_seq_start,
325 .next = ipmr_mfc_seq_next,
326 .stop = ipmr_mfc_seq_stop,
327 .show = ipmr_mfc_seq_show,
328};
329
330static int ipmr_mfc_open(struct inode *inode, struct file *file)
331{
332 return seq_open_private(file, &ipmr_mfc_seq_ops,
333 sizeof(struct ipmr_mfc_iter));
334}
335
336static struct file_operations ip6mr_mfc_fops = {
337 .owner = THIS_MODULE,
338 .open = ipmr_mfc_open,
339 .read = seq_read,
340 .llseek = seq_lseek,
341 .release = seq_release,
342};
343#endif
344
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900345#ifdef CONFIG_IPV6_PIMSM_V2
346static int reg_vif_num = -1;
347
348static int pim6_rcv(struct sk_buff *skb)
349{
350 struct pimreghdr *pim;
351 struct ipv6hdr *encap;
352 struct net_device *reg_dev = NULL;
353
354 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
355 goto drop;
356
357 pim = (struct pimreghdr *)skb_transport_header(skb);
358 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
359 (pim->flags & PIM_NULL_REGISTER) ||
360 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
Al Viroec6b4862008-04-26 22:28:58 -0700361 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900362 goto drop;
363
364 /* check if the inner packet is destined to mcast group */
365 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
366 sizeof(*pim));
367
368 if (!ipv6_addr_is_multicast(&encap->daddr) ||
369 encap->payload_len == 0 ||
370 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
371 goto drop;
372
373 read_lock(&mrt_lock);
374 if (reg_vif_num >= 0)
375 reg_dev = vif6_table[reg_vif_num].dev;
376 if (reg_dev)
377 dev_hold(reg_dev);
378 read_unlock(&mrt_lock);
379
380 if (reg_dev == NULL)
381 goto drop;
382
383 skb->mac_header = skb->network_header;
384 skb_pull(skb, (u8 *)encap - skb->data);
385 skb_reset_network_header(skb);
386 skb->dev = reg_dev;
387 skb->protocol = htons(ETH_P_IP);
388 skb->ip_summed = 0;
389 skb->pkt_type = PACKET_HOST;
390 dst_release(skb->dst);
Pavel Emelyanovdc58c782008-05-21 14:17:54 -0700391 reg_dev->stats.rx_bytes += skb->len;
392 reg_dev->stats.rx_packets++;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900393 skb->dst = NULL;
394 nf_reset(skb);
395 netif_rx(skb);
396 dev_put(reg_dev);
397 return 0;
398 drop:
399 kfree_skb(skb);
400 return 0;
401}
402
403static struct inet6_protocol pim6_protocol = {
404 .handler = pim6_rcv,
405};
406
407/* Service routines creating virtual interfaces: PIMREG */
408
409static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
410{
411 read_lock(&mrt_lock);
Pavel Emelyanovdc58c782008-05-21 14:17:54 -0700412 dev->stats.tx_bytes += skb->len;
413 dev->stats.tx_packets++;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900414 ip6mr_cache_report(skb, reg_vif_num, MRT6MSG_WHOLEPKT);
415 read_unlock(&mrt_lock);
416 kfree_skb(skb);
417 return 0;
418}
419
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900420static void reg_vif_setup(struct net_device *dev)
421{
422 dev->type = ARPHRD_PIMREG;
423 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
424 dev->flags = IFF_NOARP;
425 dev->hard_start_xmit = reg_vif_xmit;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900426 dev->destructor = free_netdev;
427}
428
429static struct net_device *ip6mr_reg_vif(void)
430{
431 struct net_device *dev;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900432
Pavel Emelyanovdc58c782008-05-21 14:17:54 -0700433 dev = alloc_netdev(0, "pim6reg", reg_vif_setup);
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900434 if (dev == NULL)
435 return NULL;
436
437 if (register_netdevice(dev)) {
438 free_netdev(dev);
439 return NULL;
440 }
441 dev->iflink = 0;
442
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900443 if (dev_open(dev))
444 goto failure;
445
Wang Chen7af3db72008-07-14 20:54:54 -0700446 dev_hold(dev);
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900447 return dev;
448
449failure:
450 /* allow the register to be completed before unregistering. */
451 rtnl_unlock();
452 rtnl_lock();
453
454 unregister_netdevice(dev);
455 return NULL;
456}
457#endif
458
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900459/*
460 * Delete a VIF entry
461 */
462
463static int mif6_delete(int vifi)
464{
465 struct mif_device *v;
466 struct net_device *dev;
467 if (vifi < 0 || vifi >= maxvif)
468 return -EADDRNOTAVAIL;
469
470 v = &vif6_table[vifi];
471
472 write_lock_bh(&mrt_lock);
473 dev = v->dev;
474 v->dev = NULL;
475
476 if (!dev) {
477 write_unlock_bh(&mrt_lock);
478 return -EADDRNOTAVAIL;
479 }
480
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900481#ifdef CONFIG_IPV6_PIMSM_V2
482 if (vifi == reg_vif_num)
483 reg_vif_num = -1;
484#endif
485
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900486 if (vifi + 1 == maxvif) {
487 int tmp;
488 for (tmp = vifi - 1; tmp >= 0; tmp--) {
489 if (MIF_EXISTS(tmp))
490 break;
491 }
492 maxvif = tmp + 1;
493 }
494
495 write_unlock_bh(&mrt_lock);
496
497 dev_set_allmulti(dev, -1);
498
499 if (v->flags & MIFF_REGISTER)
500 unregister_netdevice(dev);
501
502 dev_put(dev);
503 return 0;
504}
505
506/* Destroy an unresolved cache entry, killing queued skbs
507 and reporting error to netlink readers.
508 */
509
510static void ip6mr_destroy_unres(struct mfc6_cache *c)
511{
512 struct sk_buff *skb;
513
514 atomic_dec(&cache_resolve_queue_len);
515
516 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
517 if (ipv6_hdr(skb)->version == 0) {
518 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
519 nlh->nlmsg_type = NLMSG_ERROR;
520 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
521 skb_trim(skb, nlh->nlmsg_len);
522 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
523 rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
524 } else
525 kfree_skb(skb);
526 }
527
528 kmem_cache_free(mrt_cachep, c);
529}
530
531
532/* Single timer process for all the unresolved queue. */
533
534static void ipmr_do_expire_process(unsigned long dummy)
535{
536 unsigned long now = jiffies;
537 unsigned long expires = 10 * HZ;
538 struct mfc6_cache *c, **cp;
539
540 cp = &mfc_unres_queue;
541
542 while ((c = *cp) != NULL) {
543 if (time_after(c->mfc_un.unres.expires, now)) {
544 /* not yet... */
545 unsigned long interval = c->mfc_un.unres.expires - now;
546 if (interval < expires)
547 expires = interval;
548 cp = &c->next;
549 continue;
550 }
551
552 *cp = c->next;
553 ip6mr_destroy_unres(c);
554 }
555
556 if (atomic_read(&cache_resolve_queue_len))
557 mod_timer(&ipmr_expire_timer, jiffies + expires);
558}
559
560static void ipmr_expire_process(unsigned long dummy)
561{
562 if (!spin_trylock(&mfc_unres_lock)) {
563 mod_timer(&ipmr_expire_timer, jiffies + 1);
564 return;
565 }
566
567 if (atomic_read(&cache_resolve_queue_len))
568 ipmr_do_expire_process(dummy);
569
570 spin_unlock(&mfc_unres_lock);
571}
572
573/* Fill oifs list. It is called under write locked mrt_lock. */
574
575static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls)
576{
577 int vifi;
578
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300579 cache->mfc_un.res.minvif = MAXMIFS;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900580 cache->mfc_un.res.maxvif = 0;
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300581 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900582
583 for (vifi = 0; vifi < maxvif; vifi++) {
584 if (MIF_EXISTS(vifi) && ttls[vifi] && ttls[vifi] < 255) {
585 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
586 if (cache->mfc_un.res.minvif > vifi)
587 cache->mfc_un.res.minvif = vifi;
588 if (cache->mfc_un.res.maxvif <= vifi)
589 cache->mfc_un.res.maxvif = vifi + 1;
590 }
591 }
592}
593
594static int mif6_add(struct mif6ctl *vifc, int mrtsock)
595{
596 int vifi = vifc->mif6c_mifi;
597 struct mif_device *v = &vif6_table[vifi];
598 struct net_device *dev;
Wang Chen5ae7b442008-07-14 20:54:23 -0700599 int err;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900600
601 /* Is vif busy ? */
602 if (MIF_EXISTS(vifi))
603 return -EADDRINUSE;
604
605 switch (vifc->mif6c_flags) {
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900606#ifdef CONFIG_IPV6_PIMSM_V2
607 case MIFF_REGISTER:
608 /*
609 * Special Purpose VIF in PIM
610 * All the packets will be sent to the daemon
611 */
612 if (reg_vif_num >= 0)
613 return -EADDRINUSE;
614 dev = ip6mr_reg_vif();
615 if (!dev)
616 return -ENOBUFS;
Wang Chen5ae7b442008-07-14 20:54:23 -0700617 err = dev_set_allmulti(dev, 1);
618 if (err) {
619 unregister_netdevice(dev);
Wang Chen7af3db72008-07-14 20:54:54 -0700620 dev_put(dev);
Wang Chen5ae7b442008-07-14 20:54:23 -0700621 return err;
622 }
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900623 break;
624#endif
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900625 case 0:
626 dev = dev_get_by_index(&init_net, vifc->mif6c_pifi);
627 if (!dev)
628 return -EADDRNOTAVAIL;
Wang Chen5ae7b442008-07-14 20:54:23 -0700629 err = dev_set_allmulti(dev, 1);
Wang Chen7af3db72008-07-14 20:54:54 -0700630 if (err) {
631 dev_put(dev);
Wang Chen5ae7b442008-07-14 20:54:23 -0700632 return err;
Wang Chen7af3db72008-07-14 20:54:54 -0700633 }
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900634 break;
635 default:
636 return -EINVAL;
637 }
638
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900639 /*
640 * Fill in the VIF structures
641 */
642 v->rate_limit = vifc->vifc_rate_limit;
643 v->flags = vifc->mif6c_flags;
644 if (!mrtsock)
645 v->flags |= VIFF_STATIC;
646 v->threshold = vifc->vifc_threshold;
647 v->bytes_in = 0;
648 v->bytes_out = 0;
649 v->pkt_in = 0;
650 v->pkt_out = 0;
651 v->link = dev->ifindex;
652 if (v->flags & MIFF_REGISTER)
653 v->link = dev->iflink;
654
655 /* And finish update writing critical data */
656 write_lock_bh(&mrt_lock);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900657 v->dev = dev;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900658#ifdef CONFIG_IPV6_PIMSM_V2
659 if (v->flags & MIFF_REGISTER)
660 reg_vif_num = vifi;
661#endif
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900662 if (vifi + 1 > maxvif)
663 maxvif = vifi + 1;
664 write_unlock_bh(&mrt_lock);
665 return 0;
666}
667
668static struct mfc6_cache *ip6mr_cache_find(struct in6_addr *origin, struct in6_addr *mcastgrp)
669{
670 int line = MFC6_HASH(mcastgrp, origin);
671 struct mfc6_cache *c;
672
673 for (c = mfc6_cache_array[line]; c; c = c->next) {
674 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
675 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
676 break;
677 }
678 return c;
679}
680
681/*
682 * Allocate a multicast cache entry
683 */
684static struct mfc6_cache *ip6mr_cache_alloc(void)
685{
686 struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_KERNEL);
687 if (c == NULL)
688 return NULL;
689 memset(c, 0, sizeof(*c));
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300690 c->mfc_un.res.minvif = MAXMIFS;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900691 return c;
692}
693
694static struct mfc6_cache *ip6mr_cache_alloc_unres(void)
695{
696 struct mfc6_cache *c = kmem_cache_alloc(mrt_cachep, GFP_ATOMIC);
697 if (c == NULL)
698 return NULL;
699 memset(c, 0, sizeof(*c));
700 skb_queue_head_init(&c->mfc_un.unres.unresolved);
701 c->mfc_un.unres.expires = jiffies + 10 * HZ;
702 return c;
703}
704
705/*
706 * A cache entry has gone into a resolved state from queued
707 */
708
709static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c)
710{
711 struct sk_buff *skb;
712
713 /*
714 * Play the pending entries through our router
715 */
716
717 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
718 if (ipv6_hdr(skb)->version == 0) {
719 int err;
720 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
721
722 if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +0900723 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900724 } else {
725 nlh->nlmsg_type = NLMSG_ERROR;
726 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
727 skb_trim(skb, nlh->nlmsg_len);
728 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
729 }
730 err = rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
731 } else
732 ip6_mr_forward(skb, c);
733 }
734}
735
736/*
737 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
738 * expects the following bizarre scheme.
739 *
740 * Called under mrt_lock.
741 */
742
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300743static int ip6mr_cache_report(struct sk_buff *pkt, mifi_t mifi, int assert)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900744{
745 struct sk_buff *skb;
746 struct mrt6msg *msg;
747 int ret;
748
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900749#ifdef CONFIG_IPV6_PIMSM_V2
750 if (assert == MRT6MSG_WHOLEPKT)
751 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
752 +sizeof(*msg));
753 else
754#endif
755 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900756
757 if (!skb)
758 return -ENOBUFS;
759
760 /* I suppose that internal messages
761 * do not require checksums */
762
763 skb->ip_summed = CHECKSUM_UNNECESSARY;
764
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900765#ifdef CONFIG_IPV6_PIMSM_V2
766 if (assert == MRT6MSG_WHOLEPKT) {
767 /* Ugly, but we have no choice with this interface.
768 Duplicate old header, fix length etc.
769 And all this only to mangle msg->im6_msgtype and
770 to set msg->im6_mbz to "mbz" :-)
771 */
772 skb_push(skb, -skb_network_offset(pkt));
773
774 skb_push(skb, sizeof(*msg));
775 skb_reset_transport_header(skb);
776 msg = (struct mrt6msg *)skb_transport_header(skb);
777 msg->im6_mbz = 0;
778 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
779 msg->im6_mif = reg_vif_num;
780 msg->im6_pad = 0;
781 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
782 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
783
784 skb->ip_summed = CHECKSUM_UNNECESSARY;
785 } else
786#endif
787 {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900788 /*
789 * Copy the IP header
790 */
791
792 skb_put(skb, sizeof(struct ipv6hdr));
793 skb_reset_network_header(skb);
794 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
795
796 /*
797 * Add our header
798 */
799 skb_put(skb, sizeof(*msg));
800 skb_reset_transport_header(skb);
801 msg = (struct mrt6msg *)skb_transport_header(skb);
802
803 msg->im6_mbz = 0;
804 msg->im6_msgtype = assert;
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300805 msg->im6_mif = mifi;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900806 msg->im6_pad = 0;
807 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
808 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
809
810 skb->dst = dst_clone(pkt->dst);
811 skb->ip_summed = CHECKSUM_UNNECESSARY;
812
813 skb_pull(skb, sizeof(struct ipv6hdr));
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +0900814 }
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900815
816 if (mroute6_socket == NULL) {
817 kfree_skb(skb);
818 return -EINVAL;
819 }
820
821 /*
822 * Deliver to user space multicast routing algorithms
823 */
824 if ((ret = sock_queue_rcv_skb(mroute6_socket, skb)) < 0) {
825 if (net_ratelimit())
826 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
827 kfree_skb(skb);
828 }
829
830 return ret;
831}
832
833/*
834 * Queue a packet for resolution. It gets locked cache entry!
835 */
836
837static int
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300838ip6mr_cache_unresolved(mifi_t mifi, struct sk_buff *skb)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900839{
840 int err;
841 struct mfc6_cache *c;
842
843 spin_lock_bh(&mfc_unres_lock);
844 for (c = mfc_unres_queue; c; c = c->next) {
845 if (ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
846 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr))
847 break;
848 }
849
850 if (c == NULL) {
851 /*
852 * Create a new entry if allowable
853 */
854
855 if (atomic_read(&cache_resolve_queue_len) >= 10 ||
856 (c = ip6mr_cache_alloc_unres()) == NULL) {
857 spin_unlock_bh(&mfc_unres_lock);
858
859 kfree_skb(skb);
860 return -ENOBUFS;
861 }
862
863 /*
864 * Fill in the new cache entry
865 */
866 c->mf6c_parent = -1;
867 c->mf6c_origin = ipv6_hdr(skb)->saddr;
868 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
869
870 /*
871 * Reflect first query at pim6sd
872 */
Rami Rosen6ac7eb02008-04-10 12:40:10 +0300873 if ((err = ip6mr_cache_report(skb, mifi, MRT6MSG_NOCACHE)) < 0) {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900874 /* If the report failed throw the cache entry
875 out - Brad Parker
876 */
877 spin_unlock_bh(&mfc_unres_lock);
878
879 kmem_cache_free(mrt_cachep, c);
880 kfree_skb(skb);
881 return err;
882 }
883
884 atomic_inc(&cache_resolve_queue_len);
885 c->next = mfc_unres_queue;
886 mfc_unres_queue = c;
887
888 ipmr_do_expire_process(1);
889 }
890
891 /*
892 * See if we can append the packet
893 */
894 if (c->mfc_un.unres.unresolved.qlen > 3) {
895 kfree_skb(skb);
896 err = -ENOBUFS;
897 } else {
898 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
899 err = 0;
900 }
901
902 spin_unlock_bh(&mfc_unres_lock);
903 return err;
904}
905
906/*
907 * MFC6 cache manipulation by user space
908 */
909
910static int ip6mr_mfc_delete(struct mf6cctl *mfc)
911{
912 int line;
913 struct mfc6_cache *c, **cp;
914
915 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
916
917 for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) {
918 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
919 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
920 write_lock_bh(&mrt_lock);
921 *cp = c->next;
922 write_unlock_bh(&mrt_lock);
923
924 kmem_cache_free(mrt_cachep, c);
925 return 0;
926 }
927 }
928 return -ENOENT;
929}
930
931static int ip6mr_device_event(struct notifier_block *this,
932 unsigned long event, void *ptr)
933{
934 struct net_device *dev = ptr;
935 struct mif_device *v;
936 int ct;
937
YOSHIFUJI Hideaki721499e2008-07-19 22:34:43 -0700938 if (!net_eq(dev_net(dev), &init_net))
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900939 return NOTIFY_DONE;
940
941 if (event != NETDEV_UNREGISTER)
942 return NOTIFY_DONE;
943
944 v = &vif6_table[0];
945 for (ct = 0; ct < maxvif; ct++, v++) {
946 if (v->dev == dev)
947 mif6_delete(ct);
948 }
949 return NOTIFY_DONE;
950}
951
952static struct notifier_block ip6_mr_notifier = {
953 .notifier_call = ip6mr_device_event
954};
955
956/*
957 * Setup for IP multicast routing
958 */
959
Wang Chen623d1a12008-07-03 12:13:30 +0800960int __init ip6_mr_init(void)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900961{
Wang Chen623d1a12008-07-03 12:13:30 +0800962 int err;
963
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900964 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
965 sizeof(struct mfc6_cache),
966 0, SLAB_HWCACHE_ALIGN,
967 NULL);
968 if (!mrt_cachep)
Wang Chen623d1a12008-07-03 12:13:30 +0800969 return -ENOMEM;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900970
971 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
Wang Chen623d1a12008-07-03 12:13:30 +0800972 err = register_netdevice_notifier(&ip6_mr_notifier);
973 if (err)
974 goto reg_notif_fail;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900975#ifdef CONFIG_PROC_FS
Wang Chen623d1a12008-07-03 12:13:30 +0800976 err = -ENOMEM;
977 if (!proc_net_fops_create(&init_net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
978 goto proc_vif_fail;
979 if (!proc_net_fops_create(&init_net, "ip6_mr_cache",
980 0, &ip6mr_mfc_fops))
981 goto proc_cache_fail;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900982#endif
Wang Chen623d1a12008-07-03 12:13:30 +0800983 return 0;
984reg_notif_fail:
985 kmem_cache_destroy(mrt_cachep);
986#ifdef CONFIG_PROC_FS
987proc_vif_fail:
988 unregister_netdevice_notifier(&ip6_mr_notifier);
989proc_cache_fail:
990 proc_net_remove(&init_net, "ip6_mr_vif");
991#endif
992 return err;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +0900993}
994
Wang Chen623d1a12008-07-03 12:13:30 +0800995void ip6_mr_cleanup(void)
996{
997#ifdef CONFIG_PROC_FS
998 proc_net_remove(&init_net, "ip6_mr_cache");
999 proc_net_remove(&init_net, "ip6_mr_vif");
1000#endif
1001 unregister_netdevice_notifier(&ip6_mr_notifier);
1002 del_timer(&ipmr_expire_timer);
1003 kmem_cache_destroy(mrt_cachep);
1004}
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001005
1006static int ip6mr_mfc_add(struct mf6cctl *mfc, int mrtsock)
1007{
1008 int line;
1009 struct mfc6_cache *uc, *c, **cp;
Rami Rosen6ac7eb02008-04-10 12:40:10 +03001010 unsigned char ttls[MAXMIFS];
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001011 int i;
1012
Rami Rosen6ac7eb02008-04-10 12:40:10 +03001013 memset(ttls, 255, MAXMIFS);
1014 for (i = 0; i < MAXMIFS; i++) {
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001015 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1016 ttls[i] = 1;
1017
1018 }
1019
1020 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1021
1022 for (cp = &mfc6_cache_array[line]; (c = *cp) != NULL; cp = &c->next) {
1023 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1024 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr))
1025 break;
1026 }
1027
1028 if (c != NULL) {
1029 write_lock_bh(&mrt_lock);
1030 c->mf6c_parent = mfc->mf6cc_parent;
1031 ip6mr_update_thresholds(c, ttls);
1032 if (!mrtsock)
1033 c->mfc_flags |= MFC_STATIC;
1034 write_unlock_bh(&mrt_lock);
1035 return 0;
1036 }
1037
1038 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1039 return -EINVAL;
1040
1041 c = ip6mr_cache_alloc();
1042 if (c == NULL)
1043 return -ENOMEM;
1044
1045 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1046 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1047 c->mf6c_parent = mfc->mf6cc_parent;
1048 ip6mr_update_thresholds(c, ttls);
1049 if (!mrtsock)
1050 c->mfc_flags |= MFC_STATIC;
1051
1052 write_lock_bh(&mrt_lock);
1053 c->next = mfc6_cache_array[line];
1054 mfc6_cache_array[line] = c;
1055 write_unlock_bh(&mrt_lock);
1056
1057 /*
1058 * Check to see if we resolved a queued list. If so we
1059 * need to send on the frames and tidy up.
1060 */
1061 spin_lock_bh(&mfc_unres_lock);
1062 for (cp = &mfc_unres_queue; (uc = *cp) != NULL;
1063 cp = &uc->next) {
1064 if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1065 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1066 *cp = uc->next;
1067 if (atomic_dec_and_test(&cache_resolve_queue_len))
1068 del_timer(&ipmr_expire_timer);
1069 break;
1070 }
1071 }
1072 spin_unlock_bh(&mfc_unres_lock);
1073
1074 if (uc) {
1075 ip6mr_cache_resolve(uc, c);
1076 kmem_cache_free(mrt_cachep, uc);
1077 }
1078 return 0;
1079}
1080
1081/*
1082 * Close the multicast socket, and clear the vif tables etc
1083 */
1084
1085static void mroute_clean_tables(struct sock *sk)
1086{
1087 int i;
1088
1089 /*
1090 * Shut down all active vif entries
1091 */
1092 for (i = 0; i < maxvif; i++) {
1093 if (!(vif6_table[i].flags & VIFF_STATIC))
1094 mif6_delete(i);
1095 }
1096
1097 /*
1098 * Wipe the cache
1099 */
1100 for (i = 0; i < ARRAY_SIZE(mfc6_cache_array); i++) {
1101 struct mfc6_cache *c, **cp;
1102
1103 cp = &mfc6_cache_array[i];
1104 while ((c = *cp) != NULL) {
1105 if (c->mfc_flags & MFC_STATIC) {
1106 cp = &c->next;
1107 continue;
1108 }
1109 write_lock_bh(&mrt_lock);
1110 *cp = c->next;
1111 write_unlock_bh(&mrt_lock);
1112
1113 kmem_cache_free(mrt_cachep, c);
1114 }
1115 }
1116
1117 if (atomic_read(&cache_resolve_queue_len) != 0) {
1118 struct mfc6_cache *c;
1119
1120 spin_lock_bh(&mfc_unres_lock);
1121 while (mfc_unres_queue != NULL) {
1122 c = mfc_unres_queue;
1123 mfc_unres_queue = c->next;
1124 spin_unlock_bh(&mfc_unres_lock);
1125
1126 ip6mr_destroy_unres(c);
1127
1128 spin_lock_bh(&mfc_unres_lock);
1129 }
1130 spin_unlock_bh(&mfc_unres_lock);
1131 }
1132}
1133
1134static int ip6mr_sk_init(struct sock *sk)
1135{
1136 int err = 0;
1137
1138 rtnl_lock();
1139 write_lock_bh(&mrt_lock);
1140 if (likely(mroute6_socket == NULL))
1141 mroute6_socket = sk;
1142 else
1143 err = -EADDRINUSE;
1144 write_unlock_bh(&mrt_lock);
1145
1146 rtnl_unlock();
1147
1148 return err;
1149}
1150
1151int ip6mr_sk_done(struct sock *sk)
1152{
1153 int err = 0;
1154
1155 rtnl_lock();
1156 if (sk == mroute6_socket) {
1157 write_lock_bh(&mrt_lock);
1158 mroute6_socket = NULL;
1159 write_unlock_bh(&mrt_lock);
1160
1161 mroute_clean_tables(sk);
1162 } else
1163 err = -EACCES;
1164 rtnl_unlock();
1165
1166 return err;
1167}
1168
1169/*
1170 * Socket options and virtual interface manipulation. The whole
1171 * virtual interface system is a complete heap, but unfortunately
1172 * that's how BSD mrouted happens to think. Maybe one day with a proper
1173 * MOSPF/PIM router set up we can clean this up.
1174 */
1175
1176int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen)
1177{
1178 int ret;
1179 struct mif6ctl vif;
1180 struct mf6cctl mfc;
1181 mifi_t mifi;
1182
1183 if (optname != MRT6_INIT) {
1184 if (sk != mroute6_socket && !capable(CAP_NET_ADMIN))
1185 return -EACCES;
1186 }
1187
1188 switch (optname) {
1189 case MRT6_INIT:
1190 if (sk->sk_type != SOCK_RAW ||
1191 inet_sk(sk)->num != IPPROTO_ICMPV6)
1192 return -EOPNOTSUPP;
1193 if (optlen < sizeof(int))
1194 return -EINVAL;
1195
1196 return ip6mr_sk_init(sk);
1197
1198 case MRT6_DONE:
1199 return ip6mr_sk_done(sk);
1200
1201 case MRT6_ADD_MIF:
1202 if (optlen < sizeof(vif))
1203 return -EINVAL;
1204 if (copy_from_user(&vif, optval, sizeof(vif)))
1205 return -EFAULT;
Rami Rosen6ac7eb02008-04-10 12:40:10 +03001206 if (vif.mif6c_mifi >= MAXMIFS)
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001207 return -ENFILE;
1208 rtnl_lock();
1209 ret = mif6_add(&vif, sk == mroute6_socket);
1210 rtnl_unlock();
1211 return ret;
1212
1213 case MRT6_DEL_MIF:
1214 if (optlen < sizeof(mifi_t))
1215 return -EINVAL;
1216 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1217 return -EFAULT;
1218 rtnl_lock();
1219 ret = mif6_delete(mifi);
1220 rtnl_unlock();
1221 return ret;
1222
1223 /*
1224 * Manipulate the forwarding caches. These live
1225 * in a sort of kernel/user symbiosis.
1226 */
1227 case MRT6_ADD_MFC:
1228 case MRT6_DEL_MFC:
1229 if (optlen < sizeof(mfc))
1230 return -EINVAL;
1231 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1232 return -EFAULT;
1233 rtnl_lock();
1234 if (optname == MRT6_DEL_MFC)
1235 ret = ip6mr_mfc_delete(&mfc);
1236 else
1237 ret = ip6mr_mfc_add(&mfc, sk == mroute6_socket);
1238 rtnl_unlock();
1239 return ret;
1240
1241 /*
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001242 * Control PIM assert (to activate pim will activate assert)
1243 */
1244 case MRT6_ASSERT:
1245 {
1246 int v;
1247 if (get_user(v, (int __user *)optval))
1248 return -EFAULT;
1249 mroute_do_assert = !!v;
1250 return 0;
1251 }
1252
1253#ifdef CONFIG_IPV6_PIMSM_V2
1254 case MRT6_PIM:
1255 {
YOSHIFUJI Hideakia9f83bf2008-04-10 15:41:28 +09001256 int v;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001257 if (get_user(v, (int __user *)optval))
1258 return -EFAULT;
1259 v = !!v;
1260 rtnl_lock();
1261 ret = 0;
1262 if (v != mroute_do_pim) {
1263 mroute_do_pim = v;
1264 mroute_do_assert = v;
1265 if (mroute_do_pim)
1266 ret = inet6_add_protocol(&pim6_protocol,
1267 IPPROTO_PIM);
1268 else
1269 ret = inet6_del_protocol(&pim6_protocol,
1270 IPPROTO_PIM);
1271 if (ret < 0)
1272 ret = -EAGAIN;
1273 }
1274 rtnl_unlock();
1275 return ret;
1276 }
1277
1278#endif
1279 /*
Rami Rosen7d120c52008-04-23 14:35:13 +03001280 * Spurious command, or MRT6_VERSION which you cannot
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001281 * set.
1282 */
1283 default:
1284 return -ENOPROTOOPT;
1285 }
1286}
1287
1288/*
1289 * Getsock opt support for the multicast routing system.
1290 */
1291
1292int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1293 int __user *optlen)
1294{
1295 int olr;
1296 int val;
1297
1298 switch (optname) {
1299 case MRT6_VERSION:
1300 val = 0x0305;
1301 break;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001302#ifdef CONFIG_IPV6_PIMSM_V2
1303 case MRT6_PIM:
1304 val = mroute_do_pim;
1305 break;
1306#endif
1307 case MRT6_ASSERT:
1308 val = mroute_do_assert;
1309 break;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001310 default:
1311 return -ENOPROTOOPT;
1312 }
1313
1314 if (get_user(olr, optlen))
1315 return -EFAULT;
1316
1317 olr = min_t(int, olr, sizeof(int));
1318 if (olr < 0)
1319 return -EINVAL;
1320
1321 if (put_user(olr, optlen))
1322 return -EFAULT;
1323 if (copy_to_user(optval, &val, olr))
1324 return -EFAULT;
1325 return 0;
1326}
1327
1328/*
1329 * The IP multicast ioctl support routines.
1330 */
1331
1332int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1333{
1334 struct sioc_sg_req6 sr;
1335 struct sioc_mif_req6 vr;
1336 struct mif_device *vif;
1337 struct mfc6_cache *c;
1338
1339 switch (cmd) {
1340 case SIOCGETMIFCNT_IN6:
1341 if (copy_from_user(&vr, arg, sizeof(vr)))
1342 return -EFAULT;
1343 if (vr.mifi >= maxvif)
1344 return -EINVAL;
1345 read_lock(&mrt_lock);
1346 vif = &vif6_table[vr.mifi];
1347 if (MIF_EXISTS(vr.mifi)) {
1348 vr.icount = vif->pkt_in;
1349 vr.ocount = vif->pkt_out;
1350 vr.ibytes = vif->bytes_in;
1351 vr.obytes = vif->bytes_out;
1352 read_unlock(&mrt_lock);
1353
1354 if (copy_to_user(arg, &vr, sizeof(vr)))
1355 return -EFAULT;
1356 return 0;
1357 }
1358 read_unlock(&mrt_lock);
1359 return -EADDRNOTAVAIL;
1360 case SIOCGETSGCNT_IN6:
1361 if (copy_from_user(&sr, arg, sizeof(sr)))
1362 return -EFAULT;
1363
1364 read_lock(&mrt_lock);
1365 c = ip6mr_cache_find(&sr.src.sin6_addr, &sr.grp.sin6_addr);
1366 if (c) {
1367 sr.pktcnt = c->mfc_un.res.pkt;
1368 sr.bytecnt = c->mfc_un.res.bytes;
1369 sr.wrong_if = c->mfc_un.res.wrong_if;
1370 read_unlock(&mrt_lock);
1371
1372 if (copy_to_user(arg, &sr, sizeof(sr)))
1373 return -EFAULT;
1374 return 0;
1375 }
1376 read_unlock(&mrt_lock);
1377 return -EADDRNOTAVAIL;
1378 default:
1379 return -ENOIOCTLCMD;
1380 }
1381}
1382
1383
1384static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1385{
Rami Rosen0912ea32008-04-13 23:59:13 -07001386 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001387 return dst_output(skb);
1388}
1389
1390/*
1391 * Processing handlers for ip6mr_forward
1392 */
1393
1394static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1395{
1396 struct ipv6hdr *ipv6h;
1397 struct mif_device *vif = &vif6_table[vifi];
1398 struct net_device *dev;
1399 struct dst_entry *dst;
1400 struct flowi fl;
1401
1402 if (vif->dev == NULL)
1403 goto out_free;
1404
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001405#ifdef CONFIG_IPV6_PIMSM_V2
1406 if (vif->flags & MIFF_REGISTER) {
1407 vif->pkt_out++;
1408 vif->bytes_out += skb->len;
Pavel Emelyanovdc58c782008-05-21 14:17:54 -07001409 vif->dev->stats.tx_bytes += skb->len;
1410 vif->dev->stats.tx_packets++;
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001411 ip6mr_cache_report(skb, vifi, MRT6MSG_WHOLEPKT);
1412 kfree_skb(skb);
1413 return 0;
1414 }
1415#endif
1416
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001417 ipv6h = ipv6_hdr(skb);
1418
1419 fl = (struct flowi) {
1420 .oif = vif->link,
1421 .nl_u = { .ip6_u =
1422 { .daddr = ipv6h->daddr, }
1423 }
1424 };
1425
1426 dst = ip6_route_output(&init_net, NULL, &fl);
1427 if (!dst)
1428 goto out_free;
1429
1430 dst_release(skb->dst);
1431 skb->dst = dst;
1432
1433 /*
1434 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1435 * not only before forwarding, but after forwarding on all output
1436 * interfaces. It is clear, if mrouter runs a multicasting
1437 * program, it should receive packets not depending to what interface
1438 * program is joined.
1439 * If we will not make it, the program will have to join on all
1440 * interfaces. On the other hand, multihoming host (or router, but
1441 * not mrouter) cannot join to more than one interface - it will
1442 * result in receiving multiple packets.
1443 */
1444 dev = vif->dev;
1445 skb->dev = dev;
1446 vif->pkt_out++;
1447 vif->bytes_out += skb->len;
1448
1449 /* We are about to write */
1450 /* XXX: extension headers? */
1451 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1452 goto out_free;
1453
1454 ipv6h = ipv6_hdr(skb);
1455 ipv6h->hop_limit--;
1456
1457 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1458
1459 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev,
1460 ip6mr_forward2_finish);
1461
1462out_free:
1463 kfree_skb(skb);
1464 return 0;
1465}
1466
1467static int ip6mr_find_vif(struct net_device *dev)
1468{
1469 int ct;
1470 for (ct = maxvif - 1; ct >= 0; ct--) {
1471 if (vif6_table[ct].dev == dev)
1472 break;
1473 }
1474 return ct;
1475}
1476
1477static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache)
1478{
1479 int psend = -1;
1480 int vif, ct;
1481
1482 vif = cache->mf6c_parent;
1483 cache->mfc_un.res.pkt++;
1484 cache->mfc_un.res.bytes += skb->len;
1485
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001486 /*
1487 * Wrong interface: drop packet and (maybe) send PIM assert.
1488 */
1489 if (vif6_table[vif].dev != skb->dev) {
1490 int true_vifi;
1491
1492 cache->mfc_un.res.wrong_if++;
1493 true_vifi = ip6mr_find_vif(skb->dev);
1494
1495 if (true_vifi >= 0 && mroute_do_assert &&
1496 /* pimsm uses asserts, when switching from RPT to SPT,
1497 so that we cannot check that packet arrived on an oif.
1498 It is bad, but otherwise we would need to move pretty
1499 large chunk of pimd to kernel. Ough... --ANK
1500 */
1501 (mroute_do_pim || cache->mfc_un.res.ttls[true_vifi] < 255) &&
1502 time_after(jiffies,
1503 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1504 cache->mfc_un.res.last_assert = jiffies;
1505 ip6mr_cache_report(skb, true_vifi, MRT6MSG_WRONGMIF);
1506 }
1507 goto dont_forward;
1508 }
1509
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001510 vif6_table[vif].pkt_in++;
1511 vif6_table[vif].bytes_in += skb->len;
1512
1513 /*
1514 * Forward the frame
1515 */
1516 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1517 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1518 if (psend != -1) {
1519 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1520 if (skb2)
1521 ip6mr_forward2(skb2, cache, psend);
1522 }
1523 psend = ct;
1524 }
1525 }
1526 if (psend != -1) {
1527 ip6mr_forward2(skb, cache, psend);
1528 return 0;
1529 }
1530
YOSHIFUJI Hideaki14fb64e2008-04-03 09:22:54 +09001531dont_forward:
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001532 kfree_skb(skb);
1533 return 0;
1534}
1535
1536
1537/*
1538 * Multicast packets for forwarding arrive here
1539 */
1540
1541int ip6_mr_input(struct sk_buff *skb)
1542{
1543 struct mfc6_cache *cache;
1544
1545 read_lock(&mrt_lock);
1546 cache = ip6mr_cache_find(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1547
1548 /*
1549 * No usable cache entry
1550 */
1551 if (cache == NULL) {
1552 int vif;
1553
1554 vif = ip6mr_find_vif(skb->dev);
1555 if (vif >= 0) {
1556 int err = ip6mr_cache_unresolved(vif, skb);
1557 read_unlock(&mrt_lock);
1558
1559 return err;
1560 }
1561 read_unlock(&mrt_lock);
1562 kfree_skb(skb);
1563 return -ENODEV;
1564 }
1565
1566 ip6_mr_forward(skb, cache);
1567
1568 read_unlock(&mrt_lock);
1569
1570 return 0;
1571}
1572
1573
1574static int
1575ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm)
1576{
1577 int ct;
1578 struct rtnexthop *nhp;
1579 struct net_device *dev = vif6_table[c->mf6c_parent].dev;
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +09001580 u8 *b = skb_tail_pointer(skb);
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001581 struct rtattr *mp_head;
1582
1583 if (dev)
1584 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1585
1586 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1587
1588 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1589 if (c->mfc_un.res.ttls[ct] < 255) {
1590 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1591 goto rtattr_failure;
1592 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1593 nhp->rtnh_flags = 0;
1594 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1595 nhp->rtnh_ifindex = vif6_table[ct].dev->ifindex;
1596 nhp->rtnh_len = sizeof(*nhp);
1597 }
1598 }
1599 mp_head->rta_type = RTA_MULTIPATH;
YOSHIFUJI Hideaki549e0282008-04-05 22:17:39 +09001600 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
YOSHIFUJI Hideaki7bc570c2008-04-03 09:22:53 +09001601 rtm->rtm_type = RTN_MULTICAST;
1602 return 1;
1603
1604rtattr_failure:
1605 nlmsg_trim(skb, b);
1606 return -EMSGSIZE;
1607}
1608
1609int ip6mr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1610{
1611 int err;
1612 struct mfc6_cache *cache;
1613 struct rt6_info *rt = (struct rt6_info *)skb->dst;
1614
1615 read_lock(&mrt_lock);
1616 cache = ip6mr_cache_find(&rt->rt6i_src.addr, &rt->rt6i_dst.addr);
1617
1618 if (!cache) {
1619 struct sk_buff *skb2;
1620 struct ipv6hdr *iph;
1621 struct net_device *dev;
1622 int vif;
1623
1624 if (nowait) {
1625 read_unlock(&mrt_lock);
1626 return -EAGAIN;
1627 }
1628
1629 dev = skb->dev;
1630 if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) {
1631 read_unlock(&mrt_lock);
1632 return -ENODEV;
1633 }
1634
1635 /* really correct? */
1636 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
1637 if (!skb2) {
1638 read_unlock(&mrt_lock);
1639 return -ENOMEM;
1640 }
1641
1642 skb_reset_transport_header(skb2);
1643
1644 skb_put(skb2, sizeof(struct ipv6hdr));
1645 skb_reset_network_header(skb2);
1646
1647 iph = ipv6_hdr(skb2);
1648 iph->version = 0;
1649 iph->priority = 0;
1650 iph->flow_lbl[0] = 0;
1651 iph->flow_lbl[1] = 0;
1652 iph->flow_lbl[2] = 0;
1653 iph->payload_len = 0;
1654 iph->nexthdr = IPPROTO_NONE;
1655 iph->hop_limit = 0;
1656 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
1657 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
1658
1659 err = ip6mr_cache_unresolved(vif, skb2);
1660 read_unlock(&mrt_lock);
1661
1662 return err;
1663 }
1664
1665 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1666 cache->mfc_flags |= MFC_NOTIFY;
1667
1668 err = ip6mr_fill_mroute(skb, cache, rtm);
1669 read_unlock(&mrt_lock);
1670 return err;
1671}
1672